SUMMARY

A general sensitivity concept, namely the variance-based sensitivity analysis suited for all kind of models, is presented. The variance-based analysis is sampling-based and therefore applies Monte Carlo simulation. Moreover, it relies on the computation of conditional vari-ances. Measures that do not need a linear or additive model behavior for quantitative sensitiv-ity analysis are SOBOL’s and the FOURIER amplitude sensitivity test (FAST) sensitivity indices. The main advantage of the methods is that the analytic structure of the model to be analyzed has not to be known. Theoretically an unknown computer code may be used as model for the sensitivity analysis. The paper will deal with a kinematic KALMAN-filter used for vehicle positioning as applica-tion for the variance-based method. The state vector used for the filter algorithm consists of the coordinates of the vehicle, its velocity and its orientation. The sensitivity analysis identi-fies the influence of the measured quantities like GPS coordinates, orientation changes and distance increments on the state vector components. ZUSAMMENFASSUN